

AI & computing
Recommender system: retrieval, ranking and feedback
- Layout:
- a funnel plus a feedback loop
- Aspect ratio:
- 16:9
Prompt
Create a system diagram of 【a recommender system】, read from left to right. Draw each relationship exactly as listed and nothing more:
1. 【User behavior (small click, view and cart icons) and the item catalog (a large grid of item cards) both feed into candidate generation.】
2. 【Candidate generation is a funnel that narrows the whole catalog to a small set of candidates.】
3. 【The candidates → a ranking model (a small network) → recommendations shown on a phone screen as a few item cards.】
4. 【A dashed arrow runs from the phone back to user behavior, showing the user's new clicks as feedback.】
There are no T-bars in this figure. Do not write the numbered sentences in the figure. Write each label only once in the whole figure.
Labels, each used once, no other text: 【User behavior, Item catalog, Candidate generation, Ranking model, Recommendations, Feedback】.
Style: journal-style schematic diagram. Clean flat vector graphics on a white background. Modules are rounded rectangles or simple icons with their short names inside; light pastel panels group related layers. Only 3-4 muted colors, each with one fixed meaning. Thin, uniform arrows show the flow of data, energy, material or influence; dashed arrows for indirect or feedback links. Sans-serif English labels only, each label appears exactly once, spelled exactly as given. No 3D rendering, no shadows, no glow, no cyberpunk or tech-background decoration, no photorealism.
Opens Yila with this prompt and the example as a reference image.
The examples were generated with Nano Banana. The prompt also works with GPT Image and other image models; results vary by model.






